3dgs-visualizer

# 3DGS Visualizer — Publication-Quality Research Visualizations Generate publication-quality charts for 3DGS method landscape comparison and evolution tracking. ## Capabilities - **Radar Charts**: Multi-dimensional method capability comparison - **Comparison Tables**: Visual performance/efficiency tables with highlighting - **Method Timelines**: Chronological evolution showing trends and paradigm shifts - **Dual Output**: Static (PDF/PNG via matplotlib) and interactive HTML (via plotly) ## Data Sources | File | Content | |------|---------| | `../../references/3dgs-methods-overview.md` | Master index, metrics summary | | `../../references/methods-core.md` | Foundation, Geometry, CAD, Generation, Feed-Forward, Compression, Dynamic | | `../../references/methods-semantic-editing.md` | Semantic, Editing, Avatar, Material methods | | `../../references/methods-systems-apps.md` | Robustness, Driving, SLAM, Simulation, Cross-Domain | | `../../references/baselines.md` | Standard baselines with core metrics | | `../../references/experiments.md` | Dataset configs, efficiency reference values | --- ## Visualization 1: Radar Charts (Method Capability Comparison) **When to use**: Comparing 3–8 methods across multiple dimensions; showing quality/speed/memory trade-offs; use-case recommendation. ### Dimensions | Dimension | Scoring Criteria (0–10) | |-----------|------------------------| | **Render Quality** | 10=SOTA, 7=competitive, 5=acceptable, 3=below baseline | | **Render Speed** | 10=200+ FPS, 7=60–100, 5=30–60, 3=<30 | | **Memory Efficiency** | 10=<50MB, 7=100–500MB, 5=0.5–2GB, 3=>2GB | | **Geometry Quality** | 10=mesh-ready (2DGS/SuGaR), 7=decent depth, 5=approx, 3=poor | | **Scalability** | 10=city-scale, 7=building, 5=room, 3=object-only | | **Ease of Use** | 10=single script, 7=standard pipeline, 5=multi-stage, 3=complex setup | | **Novelty** | 10=paradigm shift, 7=significant extension, 5=incremental, 3=minor tweak | Adjust dimensions by context (compression: add "Compression Ratio"; avatar: add "Expression Fidelity"; SLAM: add "Tracking Accuracy"). ### API ```python OKABE_ITO = ['#E69F00', '#56B4E9', '#009E73', '#F0E442', '#0072B2', '#D55E00', '#CC79A7', '#000000'] # Static (matplotlib) def plot_radar(methods_data, dimensions, title="3DGS Method Comparison", output_path="radar_comparison.pdf", figsize=(8, 8)): """methods_data: {name: [score1, ...]}, dimensions: [label, ...]""" N = len(dimensions) angles = np.linspace(0, 2*np.pi, N, endpoint=False).tolist() angles += angles[:1] fig, ax = plt.subplots(figsize=figsize, subplot_kw=dict(polar=True)) for i, (name, values) in enumerate(methods_data.items()): values = values + values[:1] ax.plot(angles, values, 'o-', linewidth=2, label=name, color=OKABE_ITO[i%8]) ax.fill(angles, values, alpha=0.1, color=OKABE_ITO[i%8]) ax.set_xticks(angles[:-1]); ax.set_xticklabels(dimensions, fontsize=10) ax.set_ylim(0, 10); ax.set_yticks([2,4,6,8,10]) ax.legend(loc='upper right', bbox_to_anchor=(1.3, 1.1), fontsize=9) ax.grid(color='grey', linewidth=0.3, alpha=0.5) plt.tight_layout() plt.savefig(output_path, dpi=300, bbox_inches='tight', facecolor='white') plt.savefig(output_path.replace('.pdf','.png'), dpi=300, bbox_inches='tight', facecolor='white') plt.close() # Interactive (plotly) def plot_radar_interactive(methods_data, dimensions, title="3DGS Method Comparison", output_path="radar_comparison.html"): fig = go.Figure() for i, (name, values) in enumerate(methods_data.items()): fig.add_trace(go.Scatterpolar( r=values+values[:1], theta=dimensions+dimensions[:1], fill='toself', name=name, line_color=OKABE_ITO[i%8], opacity=0.8)) fig.update_layout(polar=dict(radialaxis=dict(visible=True, range=[0,10])), showlegend=True, title=dict(text=title), width=900, height=700) fig.write_html(output_path) ``` --- ## Visualization 2: Comparison Tables (Visual Performance Tables) **When to use**: Summarizing quantitative results across methods/datasets; paper-ready tables with visual emphasis; efficiency vs quality trade-off. ### Table Types | Type | Description | Best For | |------|-------------|----------| | **A: Quantitative Performance** | Color-coded cells (green=best, blue=second) | Multi-dataset metric comparison | | **B: Efficiency-Quality Scatter** | FPS vs PSNR scatter with category coloring | Speed/quality trade-off analysis | ### API — Type A: Performance Table ```python def plot_comparison_table(data, methods, datasets, metric="PSNR (dB)", higher_is_better=True, output_path="perf_table.pdf"): """data: 2D array [method][dataset]""" fig, ax = plt.subplots(figsize=(len(datasets)*1.8+2, len(methods)*0.6+1)) ax.axis('off') cell_text, cell_colors = [], [] for i in range(len(datasets)): row, row_colors = [], [] col_vals = [data[k][i] for k in range(len(methods))] for j in range(len(methods)): val = data[j][i]; row.append(f"{val:.2f}") is_best = abs(val - (max if higher_is_better else min)(col_vals)) < 0.01 is_second = abs(val - sorted(col_vals, reverse=higher_is_better)[1]) < 0.01 if len(col_vals)>1 else False row_colors.append('#C6EFCE' if is_best else '#BDD7EE' if is_second else '#FFFFFF') cell_text.append(row); cell_colors.append(row_colors) table = ax.table(cellText=cell_text, rowLabels=datasets, colLabels=methods, cellColours=cell_colors, loc='center', cellLoc='center') table.auto_set_font_size(False); table.set_fontsize(10); table.scale(1, 1.8) for j in range(len(methods)): table[0,j].set_facecolor('#4472C4'); table[0,j].set_text_props(color='white', fontweight='bold') ax.set_title(f"{metric} Comparison", fontsize=14, fontweight='bold', pad=20) plt.tight_layout(); plt.savefig(output_path, dpi=300, bbox_inches='tight', facecolor='white') plt.close() ``` ### API — Type B: Efficiency Scatter ```python CATEGORY_COLORS = { 'Foundation': '#0072B2', 'Compression': '#E69F00', 'Feed-Forward': '#009E73', 'Geometry': '#D55E00', 'Dynamic': '#CC79A7', 'Other': '#56B4E9', 'Surface/Geometry': '#D55E00', 'Editing': '#56B4E9', 'Semantic/Language': '#F0E442', 'Avatar/Human': '#994F00', 'SLAM': '#661100', 'Cross-Domain': '#5B5B5B', 'Robustness': '#984EA3', 'Generation': '#4daf4a', 'System/Acceleration': '#377eb8', 'CAD/Mesh': '#ff7f00', } def plot_efficiency_scatter(methods_info, output_path="efficiency_scatter.pdf"): """methods_info: [{name, psnr, fps, category, size}]""" fig, ax = plt.subplots(figsize=(8, 6)) for info in methods_info: color = CATEGORY_COLORS.get(info.get('category','Other'), '#56B4E9') ax.scatter(info['fps'], info['psnr'], s=info.get('size',100), c=color, alpha=0.8, edgecolors='black', linewidth=0.5) ax.annotate(info['name'], (info['fps'], info['psnr']), textcoords="offset points", xytext=(5,5), fontsize=8) ax.set_xlabel('Rendering Speed (FPS)'); ax.set_ylabel('PSNR (dB)') ax.axhline(y=27, color='grey', linestyle='--', alpha=0.3) ax.axvline(x=60, color='grey', linestyle='--', alpha=0.3) ax.spines['top'].set_visible(False); ax.spines['right'].set_visible(False) plt.tight_layout(); plt.savefig(output_path, dpi=300, bbox_inches='tight', facecolor='white') plt.close() # Interactive table (plotly) def plot_interactive_table(data, methods, datasets, metric="PSNR (dB)", output_path="perf_table.html"): fig = go.Figure(data=[go.Table( header=dict(values=[metric]+methods, fill_color='#4472C4', font=dict(color='white', size=12)), cells=dict(values=[[f"{v:.2f}" for v in col] for col in zip(*data)], fill_color='white'))]) fig.update_layout(width=800, title=metric); fig.write_html(output_path) ``` --- ## Visualization 3: Method Timelines (3DGS Evolution) **When to use**: Chronological development; identifying research trends; literature review figures; conference slides. ### Design Principles - **Horizontal axis**: Time (year/quarter) - **Vertical lanes**: Research categories - **Node size**: Significance (citation count) - **Node color**: Category (use CATEGORY_COLORS, consistent with other charts) - **Connections**: Show lineage (e.g., 3DGS → Scaffold-GS, 3DGS → 2DGS) ### API — Static Timeline ```python def plot_timeline(events, output_path="3dgs_timeline.pdf", figsize=(16, 10)): """events: [{name, date(YYYY-MM), category, venue, citation_count}]""" fig, ax = plt.subplots(figsize=figsize) y_positions = {cat: i for i, cat in enumerate(sorted(set(e['category'] for e in events)))} for event in events: y = y_positions[event['category']] dt = datetime.strptime(event['date'][:7], '%Y-%m') x = mdates.date2num(dt) color = CATEGORY_COLORS.get(event['category'], '#666666') size = min(200, 50 + event.get('citation_count', 20) * 0.5) ax.scatter(x, y, s=size, c=color, alpha=0.8, edgecolors='black', linewidth=0.5, zorder=5) venue = event.get('venue', '') label = f"{event['name']}\n({venue})" if venue else event['name'] ax.annotate(label, (x, y), textcoords="offset points", xytext=(0, -size**0.5/2 - 8), ha='center', fontsize=6,